Hi John,
It depends on what you think of as the null model - if the null model
is a Nested Logit with the coefficients constrained to zero and the IV
parameters constrained to unity then -clogit- gives you the correct
log-likelihood. Note that when the decision-makers face the same
number of alternatives this is simply given by N*ln(1/J), where N is
the number of choices and J the number of alternatives. In most
applications, however, it would be natural to test the unconstrained
-clogit- against this null model and then the unconstrained -nlogit-
against the unconstrained -clogit-, the former being a joint test of
the coefficients being equal to zero and the latter a joint test of
the IV parameters being equal to unity.
Hope this helps.
Arne
On 08/04/07, John Fulton <[email protected]> wrote:
Hello statalisters.
I need to get the null-model log likelihood from -nlogit-.
I can't think of a way to do it, because -nlogit's- syntax requires
nesting information,
as in this example, where the dependent variable (which indicates the
chosen alternative)
is "chosen," and the group is identified by "group_id":
. nlogit chosen (nest1=var1) (nest2=var2), group(group_id)
But I don't want the log likelihood with var1 and var2 in the model.
I tried this:
. nlogit chosen , group(group_id)
and was told "varlist not allowed"
If I do this:
. clogit chosen, group(group_id)
I get the log likelihood of the null model.
Is there any reason this isn't a valid log likelihood for the null -nlogit-?
Thank you.
John.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
